THE BELL SYSTEM
, volume xliii January 1964 number 1, part 1
Copyright 1964, American Telephone and Telegraph Company
By A. FEINER
(Manuscript received August 20, 1963)
The advantages of the fenced as a switching network cross-point led to an
early decision to adopt it for use in electronic switching systems. The
prospect of large-scale use of the device gave impetus to a search for an
economical, easily fabricated component. This paper describes the con-
siderations which influenced the choices of a suitable magnetic material,
magnetic circuit geometry, and coil design that were made for the produc-
The concept of the ferreed was presented in an earlier article in this
journal. 1 The purpose of this paper is to describe the evolution of this
device during its further development.
To recollect, a ferreed is a device born of marriage between miniature
sealed reed contacts (see Ref. 2) and an external magnetic circuit
containing remanently magnetizable members. Operation or release of
the sealed contacts can be controlled by setting the remanent members
in one of two magnetic states by means of short current pulses.
Among the several useful properties that can be brought about in
the ferreeds by selection of the proper magnetic configurations and coil
design Is the ability to respond to coordinate excitation — a vital re-
quirement for any device considered for a network crosspoint.
Recognition of the potential advantages of a switching network cross-
point with metallic contacts, absence of holding power and the ability
to operate in times much shorter than prior electromechanical devices
2 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1964
led to an early decision to adopt it for the network of No. 1 ESS (Elec-
tronic Switching System) — the new telephone switching system sched-
uled for its commercial debut in 1965.
The intended application of the ferreed in the switching network of
No. 1 ESS, where it would appear in very large numbers (14-20 cross-
points per line), gave impetus to a search for an economical, easily
fabricated embodiment. Several important choices had to be made with
regard to the geometry of the magnetic circuit, the winding configuration
and the remanent magnetic material. At the same time, the require-
ments of the sealed reed contact were reexamined, and a modified ver-
sion of it known as the 237B contact was adopted for ferreed use.
II. THE CROSSPOINT FERREED
2.1 Choice of Remanent Material
All original work on the ferreeds was based on the use of a specially
developed cobalt ferrite as the remanent material. In time, certain
inherent difficulties became apparent: notably, a strong temperature
dependence of the magnetic properties and low flux density, leading to
structures of large cross section and poor efficiency. Furthermore, as
more thought was given to the ferreed as a system component, it was
found that the originally postulated microsecond speeds for the actuation
of the ferreed were neither required nor practical from the standpoint
of driving requirements.
These considerations opened the way to a search for a metallic sub-
stitute. Several chromium and tungsten steel compositions were investi-
gated and found wanting due to lack of squareness and fullness of the
hysteresis loop — ■ properties whose importance were stressed in Ref . 1 .
The attention soon centered on a recent addition to the list of cobalt-
iron- vanadium alloys — Remendur. The name of this alloy refers to its
primary magnetic characteristic, i.e., a remanence greater than 17,000
gauss. This is coupled with a square hysteresis loop and a coercive force
from 1 to 60 oersteds. With a nominal composition of 48 per cent cobalt,
48 per cent iron, 3.5 per cent vanadium and 0.5 per cent manganese,
Remendur bridges the gap between the high coercive force of Vicalloy
and the low coercive force and high permeability properties of 2V-
Permendur and Supermendur. Fig. 1 shows a hysteresis loop obtained
on a Remendur strip developed for ferreed use. Of import ance to the
ferreed application is the squareness B r /B s and fullness \ZH B /H c B r
H IN OERSTEDS
Fig. 1 — Hysteresis loop of Remendur used in ferreeds.
of the hysteresis loop. This property implies that the energy expendi-
ture in establishing a desired end state approaches a minimum, and
that the excess flux generated in the same process is small — important
in view of the interference problems present in ferreed arrays.
2.2 Choice of Geometry
There exist two basic forms of ferreed structures — the parallel and
the series ferreeds. These are illustrated in Fig. 2. The choice of Remen-
dur, the need for tight magnetic coupling between the remanent mem-
bers and the reed contacts, and the relative ease of fabrication led to
adoption of the series structure for the crosspoint ferreed.
That structure is shown in Fig. 3 in the form used in the ESS network.
Mounted on each side of the reed contacts, which are molded together
in plastic to form a single piece part, and extending approximately over
the length of the glass envelopes, are two flat plates of Remendur.
Notches on the plastic and on the plates permit accurate relative posi-
tioning of the two.
The reeds and the remanent plates are inserted into plastic coil forms
THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1964
Fig. 2 — Principles of parallel and series ferreeds.
molded into a steel plate. This steel plate acts as a common shunt for
the whole array — it divides each crosspoint magnetically into two
separately controllable halves, greatly reducing the energy requirement
for producing the release state in which, as shown in Fig. 4, the two
halves of the remanent members are magnetized in opposing directions.
The same steel plate acts as the mechanical backbone of the whole
2.3 Coil Design
The differential excitation mode was selected to provide coordinate
addressing of crosspoints. Fig. 5 reviews this principle as applied to a
series ferreed. Each crosspoint has two sets of windings — one for each
coordinate. Each set contains a winding of N turns on one side of the
shunt plate and one with a larger number, typically 2N, on the other
side. The 2iV-turn winding is connected series opposing the A^-turn
winding. One pair of windings is in series with the corresponding pairs
of all crosspoints in the same row, while the other is in series with the
pairs of all crosspoints in the same column of the array. As the paired
windings oppose each other, energization produces the release state in
every crosspoint energized, except the one where both pairs of windings
THE FERREED 5
are energized simultaneously — the crosspoint at the intersection of the
energized row and the column.
The logic inherent to differential excitation was found to be well
suited to network array operation, in which, in general, only one cross-
point in each row or column need be operated.
No separate release actions are required, as operating a crosspoint
automatically releases other crosspoints associated with the same row
The design of the coils has to take in account the energization re-
quirements of a single crosspoint as well as the system requirement
237 B SEALED CONTACTS
2 PER MOLDED ASSEMBLY
Fig. 3 — Exploded view of the two-wire crosspoint ferreed.
THE BELT, SYSTEM TECHNICAL JOURNAL, JANUARY 1904
<j in ni — - <\j o
saaisuBO ni Nouoauia ivixv
3H± NI A1ISN31NI 01313 DI13N3VW
J I L
^ fi) (u - - ni pi t •o <o
sa3isa3o ni Nouosaia ~ivixv
3H1 NI A1ISN31NI Q13I3 0I13N3VW
K = 2
Fig. 5 — Winding configuration for differential excitation of the series ferreed:
(a) winding pattern, (b) mirror symbol notation.
calling for simultaneous pulsing of 32 winding pairs in the process of
establishing a connection through two stages of ferreed switches.
In ESS, these considerations led to the adoption of coils with windings
of 18 and 39 turns wound with 25-gauge copper wire. With these coils,
the nominal operating current pulse of 10 amperes peak amplitude and
250 microseconds duration insures adequate margins for both operation
and release of the crosspoint.
The coils are wound directly on the coil forms by a machine that
winds eight rows (or columns) of crosspoints simultaneously in a con-
tinuous succession, each with a single length of wire. This eliminates
THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1964
soldered connections between coils, thus reducing the winding cost and
improving the reliability of the assembly.
The winding sense is reversed in adjacent crosspoints. This magnetic
"checkerboarding" was found to be an effective means for reducing
magnetic interaction phenomena as well as the noise pickup in the
transmission pairs due to ferreed energizing pulses.
2.4 Crosspoint Arrays
Switching network considerations led to selection of an 8 X 8 cross-
point array as a basic network building block. In Fig. 6, such an array
is shown. In addition, specifically for the concentrating stages of the
network, several other array types were required: a switch providing
each of 16 input terminal pairs with an access to 4 out of 8 available
outputs, and 8X4 and 4X4 switches. It was found that each of these
arrays could be derived from the basic 8X8 apparatus unit by suitably
changing the connections of the control windings and the voice-pair
strappings. Fig. 7 shows these connections for all the developed ferreed
Fig. 6 — An 8 X 8 ferreed switch with covers removed.
switch types. As can be expected, this standardization of the physical
size and component parts of the switches has eased the manufacturing
and the network equipment design problems.
The connections shown between the ends of the row and column
control winding chains stem from the access scheme adopted in the
network design. In this scheme, identical current is applied to both
coordinates by connecting them effectively in series when energizing a
crosspoint at their intersection.
\ — I T
1 1 I i-
T T T T
Fig. 7 — Control winding interconnection for three types of two-wire switches:
(a) 16 X 4/8, (b) 8X4, and (c) 8X8.
10 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1964
III. DESIGN TECHNIQUE
When the problem of designing the ferreed was first approached, it
was found that the usual lumped-constant, linear magnetic circuit
approach, while sufficient to yield a workable device, did not provide
the means for its optimization; neither did it give an assurance of
margins in face of tolerance allowances that have to be made for the
whole structure, and variations in reed contact properties and in the
magnetic properties of Remendur. Several attempts were made to refine
the analytical tools toward this end. While providing qualitative in-
sight into the operation of the device, they were frustrated from attain-
ing the ultimate goal of a quantitative, explicit solution by the complex-
ity of the problem caused by the rather difficult geometry and the
essential nonlinearity of the magnetic materials.
As a result, the refinements in the ferreed design had to be based
largely on experimental techniques. Over the years, numerous experi-
mental ferreed study techniques have been devised. These include the
use of search coils with integrators, hysteresis measurements of reeds
and the remanent magnetic members, Hall probes in the crosspoint
structure and the reed gap, and reversible permeability measurements
of the reeds. Supplemented by experiments in which the component
parts of the structure, their positioning and the driving conditions
underwent systematic variations, these techniques were instrumental
in arriving at the present structure.
The use of Hall probes provided two study techniques. First, Hall
probes were employed to measure longitudinal magnetic field intensity
along the ferreed axis, after applying varying operate and release pulses.
Second, via the use of specially constructed sealed reeds with Hall
probes mounted in the gap of the reed, it was possible to measure the
resultant magnetic flux density in the reed gap under varying operating
conditions. The drawback of the techniques lies in the upsetting of the
ferreed magnetic circuit by the absence of the reed or introduction of a
permanently open reed structure.
Reversible permeability measurements of the sealed reeds, accom-
plished via inductance measurements of small sense coils at about 100
kc, provided a convenient means of determining the instantaneous ap-
plied mmf to the sealed reeds under varying operating and interference
conditions. The technique was especially useful because it permitted
the use of ordinary sealed reeds under actual operating conditions, and
it was free of drift problems since no integrator circuits were involved.
On the other hand, the nature of the reversible permeability character-
THE FERREED 11
istic of the sealed reed is so insensitive in the released state of the sealed
reeds as to make its use not suitable in that region.
IV. OTHER FERREED TYPES
4.1 The Bipolar Ferreed
In the process of designing a ferreed switching network, the need
arose for a device containing a pair of contacts that would be indi-
vidually controllable. A typical use for this device is disconnection of the
line current sensing element at the line circuit whenever a connection
is established in the switching network (cutoff relay function) . A postu-
lated property of this device — to respond to control current pulse polarity
to open or close its contacts — was found to permit integrating the con-
trol access with the one for the crosspoints.
An adaptation of the parallel ferreed principle, shown in Fig. 8,
provided a suitable embodiment meeting this need. Of the two parallel
remanent members, one consists of a permanent magnet material,
Cunifc I; the other, surrounded by a single coil, of Remendur. Con-
tact closure or release depends on the polarity of the current pulse
applied to the coil. Eight such devices packaged together form a single
apparatus unit compatible in its length with the crosspoint units.
4.2 The Four-Wire Crosspoint Array
For use in switching networks requiring two separate directions of
transmission, the two-wire crosspoint design has been extended to
permit the operation of four contacts at every crosspoint location. The
four contacts are arranged in a square pattern and are surrounded by an
open-ended box formed by four remanent plates. The windings are
similar to those of the two-wire array and again an eight-by-eight size
has been chosen; Fig. 9 shows an individual crosspoint and an over-
all view of the unit.
Out of the original concept of the ferreed originated a whole class of
useful switching devices. Characterized by small size, high speed of
operation and absence of holding power, they permit retaining the
desirable aspects of metallic contacts in the environment of electronic
switching machines without creating undue time compatibility problems.
X SEALED CONTACT
Fig. 8 — (a) The bipolar ferreed; (b) a 1 X 8 apparatus unit.
2 PER MOLDED
Fig. !) — (a) Exploded view of a single four-wire crosspoint; (b) over-all view
of an 8X8 switch with protective covers removed.
THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1964
Table I — Summary of Ferreed Characteristics
2-wire (2) 8 X 4
2-wire 16 X 4/8
2 X 10«f
2-wire 1 X 8
2 X 10 a
* To protect the contacts, crosspoints are operated and released in a dry cir-
cuit — maximum surge current refers to current value applied to closed contacts,
t Minimum life of 2 X 10 6 operations with contact resistance below 0.2 ohm.
X This contact breaks a maximum of 40 ma in its operation.
Table I gives a summary of the characteristics of the ferreed codes how
Many people have contributed important ideas and skills to make the
ferreed a success ; the author would like to offer his particular apprecia-
tion to Messrs. H. L. B. Gould and D. H. Wenny for their work on the
Remendur, Messrs. R. L. Peek, F. H. Myers, and H. Raag for their
work in magnetic design of the ferreed, and Messrs. H. J. Wirth and
R. A. Billhardt for the mechanical design.
The credit for solving the manufacturing problems should go to Mr.
G. A. Mitchell of the Western Electric Company at Columbus.
1. Feiner, A., Lovell, C. A., Lowry, T. N., and Ridinger, P. G., The Ferreed —
A New Switching Device, B. S.T.J. , 39, January, 1960, p. 1.
2. Keller, A. C, Recent Developments in Bell System Relays — Particularly
Sealed-Contact and Miniature Relays, B. S.T.J. , this issue, p. 15.